ibmcom/powerai repository overview

⁠Requirements

PowerAI is optimized to leverage the unique capabilities of IBM Power Systems accelerated servers and x86_64 platform servers. It is supported on:

• IBM AC922 POWER9 system with NVIDIA Tesla V100 GPUs
• IBM S822LC POWER8 system with NVIDIA Tesla P100 GPUs
• IBM Power System S822LC with NVIDIA Tesla P100 GPUs
• x86_64 systems with NVIDIA Tesla V100, P100, or T4 GPUs

⁠Host System Requirements

*All PowerAI images starting with 1.5.4 require nvidia-docker 2.0

• Configuring host for PowerAI

  To setup your machine for use with PowerAI please follow instructions described in IBM's Knowledge Center for Red Hat or Ubuntu here⁠

  The instructions include how to install the NVIDIA GPU driver, docker, nvidia-docker etc.

⁠Using the PowerAI image from Docker Hub

To start up a PowerAI container.

⁠Red Hat docker

docker run -ti --env LICENSE=yes ibmcom/powerai:<tag> bash 

⁠docker-ce

nvidia-docker run -ti --env LICENSE=yes ibmcom/powerai:<tag> bash 

  PyTorch Users: If you plan on using any multiprocessor data loader with PyTorch. The default shared memory segment size for the container may not be large enough. You can increase the shared memory size with either --ipc=host or --shm-size command line options on docker run

You can read more about this issue on PyTorch's Readme https://github.com/pytorch/pytorch/blob/master/README.md⁠ under the "Docker image" section.

⁠License Acceptance

You must accept the licenses of all included components before using a PowerAI container. View the WML CE license locally in the image at $HOME/.powerai/powerai-license/1.7.0/license# or View the WML CE license externally at https://github.com/IBM/powerai/tree/master/containers/1.7.0⁠

• Accept the licenses at docker runtime by adding the --env LICENSE=yes parameter on the docker run command line

or

• Accept the license within an already running container accept-powerai-license.sh

⁠Tag Formats

• latest - PowerAI 1.7.0 and Anaconda for python 3.7

• <powerai-version>-<framework>-<cpu>-<OS>-<python>-<architecture>

  • powerai-version - The version of PowerAI installed in image latest is 1.7.0
    Available options - 1.5.2, 1.5.3, 1.5.4, 1.6.0, 1.6.1 1.6.2 1.7.0

  • framework - The framework installed on the image, (options vary depending on image version)
    Available options - all, tensorflow(>=1.6.0), tensorflow-serving (>=1.6.1), pytorch(>=1.6.0), caffe(>=1.6.0), snap-ml(>=1.6.0, ppc64le only, >=1.7.0, x86_64 and ppc64le), xgboost(>=1.6.1, ppc64le only, >=1.6.2, x86_64 and ppc64le), rapids (>=1.6.2, ppc64le only)

  • cpu - Starting with PowerAI 1.6.1, additional images are built with cpu only versions of deep learning frameworks. These versions have the -cpu flag appended to the framework name.
    Available options - all-cpu(>=1.6.1), tensorflow-cpu(>=1.6.1), caffe-cpu(>=1.6.1), xgboost-cpu(>=1.6.1, ppc64le only, >=1.6.2, x86_64 and ppc64le), pytorch-cpu(>=1.6.2), tensorflow-serving-cpu(>=1.7.0)

  • os - The operating system installed on the image (options vary depending on image version)
    Available options - ubuntu16.04(<1.6.0), ubuntu18.04(>=1.6.0)

  • python - The python version used by frameworks
    Available options - py3(<=1.6.1), <none>(python2, (<=1.6.1)), py36(>=1.6.2), py37 (>=1.6.2)

  • architecture - Starting with PowerAI 1.6.0, images are built to support x86_64 and ppc64le architectures (options vary depending on image version)
    Available options - ppc64le(>=1.6.0), x86_64(>=1.6.0), <none>(>=1.6.0 docker will auto detect on your box, <1.6.0 docker will serve ppc64le versions)

  ⁠examples

  1.7.0-pytorch-ubuntu18.04-py36 #Download pytorch image for the requesting machine's architecture with python 3.6

  1.7.0-all-ubuntu18.04-py37 #Download all frameworks image for the requesting machine's architecture with python 3.7

⁠Installed Packages

PowerAI provides software packages for several Deep Learning frameworks, supporting libraries, and tools:

⁠Container Usage

⁠User Ids

To satisfy our cloud users, and to stay inline with the principle of least privilege, the default user is pwrai in the images.

pwrai has a uid:gid of 2051:2051 and has password-less sudo setup.

For those that wish to use root, it is enabled, and available via the docker --user root argument at runtime.

For security reasons, we recommend you take advantage of Docker namespaces as documented here⁠

⁠Running PowerAI Frameworks

Please reference the "Getting Started with MLDL Frameworks" page here⁠

⁠Using TensorFlow-Serving container

The PowerAI TensorFlow-Serving docker container is modeled after the tensorflow/serving⁠ docker container provided by TensorFlow. As such all TensorFlow documentation⁠ for TensorFlow Serving with Docker applies to the PowerAI TensorFlow Serving image. Just replace tensorflow/serving in all examples with ibmcom/powerai:<powerai-version>-tensorflow-serving-ubuntu18.04-<py36/py37> or ibmcom/powerai:<powerai-version>-tensorflow-serving-cpu-ubuntu18.04-<py36/py37>

⁠LICENSES

• NVIDIA License⁠
• Anaconda License⁠
• Ubuntu License⁠
• PowerAI License⁠

⁠NVIDIA

• CUDA Toolkit To view the license for the CUDA Toolkit included in this image, click here⁠

• CUDA Deep Neural Network library (cuDNN) To view the license for cuDNN included in this image, click here⁠

⁠Anaconda

The Anaconda User's license can be viewed at (https://docs.anaconda.com/anaconda/eula⁠)

The list of installed python packages under anaconda can be displayed using pip show <packagename> | grep License:

To view a python package's specific License, go to the package's website displayed by the pip show <packagename> | grep Home-page:

⁠Ubuntu

Ubuntu's(Canonical) Legal information can be viewed at (https://www.ubuntu.com/legal⁠)

The list of installed Debian packages can be seen using dpkg --list

The license of a particular Debian package can be viewed inside the PowerAI image under /usr/share/<packagename>/copyright

⁠PowerAI

View the WML CE license locally in the image at $HOME/.powerai/powerai-license/1.7.0/license# or View the WML CE license externally at https://github.com/IBM/powerai/tree/master/containers/1.7.0⁠